fix: Timmy startup crashes and clean initialization

- Remove show_tool_calls kwarg (not in Agno 2.5.3), which crashed Agent.__init__
- Guard memory_search against top_k=None from model, return formatted string
- Skip Telegram/Discord startup silently when no token configured
- Replace placeholder MEMORY.md with proper structured hot memory document

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

MEMORY.md

@@ -1 +1,71 @@
-Good morning, I hope you had a great night.
+# Timmy Hot Memory
+
+> Working RAM — always loaded, ~300 lines max, pruned monthly
+> Last updated: 2026-02-26
+
+---
+
+## Current Status
+
+**Agent State:** Operational
+**Mode:** Development
+**Model:** llama3.2 (local via Ollama)
+**Backend:** Ollama on localhost:11434
+**Dashboard:** http://localhost:8000
+
+---
+
+## Standing Rules
+
+1. **Sovereignty First** — No cloud AI dependencies
+2. **Local-Only Inference** — Ollama on localhost
+3. **Privacy by Design** — Telemetry disabled
+4. **Tool Minimalism** — Use tools only when necessary
+5. **Memory Discipline** — Write handoffs at session end
+6. **Clean Output** — Never show JSON, tool calls, or function syntax
+
+---
+
+## System Architecture
+
+**Memory Tiers:**
+- Tier 1 (Hot): This file (MEMORY.md) — always in context
+- Tier 2 (Vault): memory/ directory — notes, profiles, AARs
+- Tier 3 (Semantic): Vector search over vault content
+
+**Swarm Agents:** Echo (research), Forge (code), Seer (data)
+**Dashboard Pages:** Briefing, Swarm, Spark, Market, Tools, Events, Ledger, Memory, Router, Upgrades, Creative
+
+---
+
+## Agent Roster
+
+| Agent | Role | Status |
+|-------|------|--------|
+| Timmy | Core AI | Active |
+| Echo | Research & Summarization | Active |
+| Forge | Coding & Debugging | Active |
+| Seer | Analytics & Prediction | Active |
+
+---
+
+## User Profile
+
+**Name:** (not set)
+**Interests:** (to be learned)
+
+---
+
+## Key Decisions
+
+(none yet)
+
+---
+
+## Pending Actions
+
+- [ ] Learn user's name and preferences
+
+---
+
+*Prune date: 2026-03-25*

src/dashboard/app.py

@@ -121,15 +121,21 @@ async def lifespan(app: FastAPI):
     if spark_engine.enabled:
         logger.info("Spark Intelligence active — event capture enabled")
 
-    # Auto-start Telegram bot if a token is configured
+    # Auto-start chat integrations (skip silently if unconfigured)
     from telegram_bot.bot import telegram_bot
-    await telegram_bot.start()
-
-    # Auto-start Discord bot and register in platform registry
     from chat_bridge.vendors.discord import discord_bot
     from chat_bridge.registry import platform_registry
     platform_registry.register(discord_bot)
-    await discord_bot.start()
+
+    if settings.telegram_token:
+        await telegram_bot.start()
+    else:
+        logger.debug("Telegram: no token configured, skipping")
+
+    if settings.discord_token or discord_bot.load_token():
+        await discord_bot.start()
+    else:
+        logger.debug("Discord: no token configured, skipping")
 
     yield
 

src/timmy/agent.py

@@ -138,7 +138,6 @@ def create_timmy(
         num_history_runs=20,
         markdown=True,
         tools=[tools] if tools else None,
-        show_tool_calls=False,
         telemetry=settings.telemetry_enabled,
     )
 

src/timmy/semantic_memory.py

@@ -319,6 +319,25 @@ semantic_memory = SemanticMemory()
 memory_searcher = MemorySearcher()
 
 
-def memory_search(query: str, top_k: int = 5) -> list[tuple[str, float]]:
-    """Simple interface for memory search."""
-    return semantic_memory.search(query, top_k)
+def memory_search(query: str, top_k: int = 5) -> str:
+    """Search past conversations and notes for relevant context.
+
+    Args:
+        query: What to search for (e.g. "Bitcoin strategy", "server setup").
+        top_k: Number of results to return (default 5).
+
+    Returns:
+        Formatted string of relevant memory results.
+    """
+    # Guard: model sometimes passes None for top_k
+    if top_k is None:
+        top_k = 5
+    results = semantic_memory.search(query, top_k)
+    if not results:
+        return "No relevant memories found."
+    parts = []
+    for content, score in results:
+        if score < 0.2:
+            continue
+        parts.append(f"[score {score:.2f}] {content[:300]}")
+    return "\n\n".join(parts) if parts else "No relevant memories found."

tests/test_agent.py

@@ -267,8 +267,8 @@ def test_create_timmy_includes_tools_for_large_model():
         assert kwargs["tools"] == [mock_toolkit]
 
 
-def test_create_timmy_show_tool_calls_false():
-    """show_tool_calls should always be False to prevent raw JSON in output."""
+def test_create_timmy_no_show_tool_calls():
+    """show_tool_calls must NOT be passed — Agno 2.5.3 doesn't support it."""
     with patch("timmy.agent.Agent") as MockAgent, \
          patch("timmy.agent.Ollama"), \
          patch("timmy.agent.SqliteDb"):
@@ -277,4 +277,4 @@ def test_create_timmy_show_tool_calls_false():
         create_timmy()
 
         kwargs = MockAgent.call_args.kwargs
-        assert kwargs["show_tool_calls"] is False
+        assert "show_tool_calls" not in kwargs